Evaluating isolation methods for human gut viral communities

Characterization of gut viral communities is of great importance: there are enormous viruses living in human gut – equaling or exceeding the number of human cells, whereas the majority of this population remain uncharacterized. Molecular and computational methods have both been utilized in previous studies to isolate and quantify the viral portion of microbial communities.
These have ranged from experimental approaches that isolate virus-like particles (VLPs), to the analysis platforms profiling the compositions and functions of viral communities. However, these studies mostly focused on analyzing the resulting viral communities without systematic evaluations of the isolation strategy. Particularly, it often remains unclear whether the selected protocol could efficiently remove bacterial-originated nucleic acids while leaving the viral potion unaffected.
Here, we present our work in optimizing and validating a workflow for the isolation of virus-like particles from gut microbial communities with thorough evaluation of different experimental parameters known to potentially affect viral extraction, including storage buffer, filtration combinations, methods for nucleic acid concentration after filtration, and various enzymatic treatments. Different combinations of the above parameters were first evaluated in stool samples spiked with simple synthetic viral communities comprising an sRNA phage (MS2) and a dsDNA phage (T4), to find out a combination with the highest purification efficiency as well as the minimum impact on the spiked viruses. The optimized protocol was then applied in more complicated mock viral communities consisting of more complexed viruses representing several common viral families in the gut (i.e. T4, T1, P1, PhiX174 and MS2). Stool samples were spiked with these mock communities, to see if the VLP isolation protocol caused any differential bias to different families of viruses. Results showing that the optimized VLP isolation protocol efficiently reduces the signals from bacterial communities (16S rRNA genes) from ~105 copies/ml to undetectable in mock communities. By comparison, minimal variations were observed in virus copies between whole communities vs. VLP concentrated communities. When applied in stool samples spiked with synthetic viral communities, the protocol depleted bacterial signals by approximately 100 fold. P1, T1 and T4 phages were depleted in VLP concentrated communities, while PhiX174 and MS2 phages were slightly enriched. This differentiation is likely introduced by the detection method (i.e. qPCR) rather than the protocol itself, given that the quantification of viruses varied greatly across different measurements.
We are currently continuing this work to further validate this VLP isolation protocol using shotgun metagenomic and metatranscriptomic sequencing in spiked stool samples (as above) and in stool samples collected from preemie babies, as well as to develop optimized computational approaches for the taxonomic and functional profiling of gut viral communities. These can circumvent some of the current limitations of the VLP isolation protocol, providing a complementary view of viral communities, and directly observing functional activities involved in the gut-virome-bacteriome interaction.